Installation

(Recommended) You can install the latest version of PyCxsim with this command:

pip install git+https://github.com/Aatamte/PyCxsim.git

An (outdated) version is also hosted on Pypi:

pip install pycxsim

Overview

PyCxsim is a framework to simulate computational agents in a confined environment.

from cxsim import Environment
from cxsim.artifacts import Marketplace
from cxsim.agents import Agent

# defining an environment
cxenv = Environment()

# adding an Artifact to the environment
market = Marketplace()
cxenv.add(market)

#adding an agent to the environment
agent = Agent()
cxenv.add(agent)

#the simulation loop
for episode in cxenv.iter_episodes():
    # start the next episode
    cxenv.reset()
    
    for step in cxenv.iter_steps():
        # start the next step
        
        for agent in cxenv.agents:
          cxenv.execute(agent, "action")
          
        cxenv.step()

(TODO) Building Agents

Multimodal

1. Text
2. Vision

GUI

One of the unique (and cool!) features of Pycxsim is the embedded GUI. For now, the GUI is being rendered with dearpygui.

Examples

1. Simulate a marketplace made up of computational agents, based off experiments in the paper “An Experimental Study of Competitive Market Behavior” by Vernon Smith.

from cxsim.examples import Smith1962Environment
import openai
import os

openai.api_key = os.environ["openai_api_key"]

cxenv = Smith1962Environment(n_agents=10, model_id="gpt-4")
cxenv.test_one(market_depth=10)

Standard Artifacts

Below are the standard artifacts provided with the CAES package:

• Marketplace
  • Agents can trade goods with each other (capital <-> good transactions only)
• Dialogue
  • Messaging
• Gridworld
  • Agents can move around the map

(TODO) Documentation

Documentation.